Abstract:

Research background: Business performance measurement is currently an important process, especially when evaluating the success of businesses and their competitiveness and calculating their market value. Business performance measurement (BPM) systems have seen an increase in both their use and popularity over the past twenty years. In these systems, financial indicators for measuring business performance are particularly important. Several approaches or frameworks have been developed for building and managing BPM systems..

Purpose of the article: The aim is to find out which financial features are the performance determinants for the construction industry. Here, a research gap can be seen, as the performance of a large sample of businesses was examined, allowing for significant generalization of results for the given industry using ensemble classifiers applied only in a few studies.

Methods: The selection of financial performance determinants for construction industry is based on a two-step procedure. In the first step, significant financial features are selected using the Elastic net algorithm. The second step of features selection is performed using Decision tree (DT), Gradient boosted tree (GBT) and AdaBoosted tree (ABT).

Findings & Value added: The results confirm that DT, GBT and ABT with features selection using elastic net achieve high classification accuracy. Results of these models outperform results of DT, GBT and ABT models with features selection using Lasso, as well as results of other machine learning models built in previous studies. The benefit of this study is the selection of most significant performance determinants for construction industry - Return on sales, Working capital to total assets, Current ratio and Total debt to total assets.